Although personality disorders (PDs) and subthreshold personality pathology are increasingly implicated as diatheses and moderators of treatment response for many Axis I disorders, researchers have identified a number of problems that limit their utility, two of which are particularly significant. First, dimensional approaches to personality diagnosis consistently outperform categorical diagnosis, and it appears clear that DSM-V will move toward dimensional diagnosis of PDs. However, no study has yet examined the comparative validity of the major dimensional alternatives. Second, virtually all PD research relies on patient self-reports; however, increasing evidence suggests limitations in the exclusive reliance on self-reports. The primary aim of this proposal is to evaluate the comparative validity of categorical PD diagnosis and four dimensional alternatives: dimensionalized DSM-IV diagnosis (number of symptoms met per disorder); the recently developed consensus four-factor trait model identified factor-analytically from self-report data; dimensional PD syndromes derived empirically; and PD traits derived factor-analytically using a method that does not rely primarily on self-reports. The secondary aim, given the potential retention in DSM-V of a dimensionalized version of the current PDs, is to compare methods of diagnosing patients dimensionally using the current PDs that vary in their reliance on self-reports. Ss will be 240 patients across three sites. For both aims, we will use the following criterion variables to assess validity: (a) aggregated lay informant-report PD and trait data; (b) treating-clinician PD and trait data; (c) adaptive functioning assessed concurrently and longitudinally (18-month follow-up); (d) etiologic data, including developmental history (childhood trauma), family history of psychiatric disorders, and molecular genetics; and (e) laboratory (implicit) measures targeted to assess specific PDs. Our general goal is to assess validity using two sets of criteria that today are largely seen as a single approach: Robins and Guze criteria (i.e., whether diagnoses differ in etiology, prognosis, laboratory tests, etc.) and construct validation (i.e., whether diagnoses differ in their external correlates in predictable ways).